Machine Learning - Research (EN)

NucleusResearch.com 2 Document Number: T176 December 2019 O V E R V I E W Most companies simply do not have the available developer talent, data science expertise, and computing infrastructure to internally develop machine learning capabilities. Because of this, interest in the field and educational opportunities have grown, along with online documentation and optimized code libraries that lower the barrier to entry for undertaking machine learning initiatives and expanded the talent pool. Amazon SageMaker is a fully managed cloud service that allows users to build, train, and deploy machine learning models at scale. SageMaker includes the most commonly employed machine learning algorithms, pre-built and optimized for use out-of-the-box. SageMaker provides high-performance, scalable machine learning algorithms, optimized for speed, scale, and accuracy, that can perform training on petabyte-scale data sets. Training models is compute-heavy and can be time-consuming and expensive to run on machinery in-house. In Amazon SageMaker, models can be automatically trained and optimized for accuracy at any scale. The process runs on AWS-managed infrastructure, so internal resources are free for other tasks, and time isn't wasted waiting for models to train on physical computers on-site. Users can deploy trained and optimized models on auto-scaling Amazon EC2 clusters spread across multiple availability zones in accordance with data usage regulations and company locations. Since its release in 2017, Amazon has released a number of additional modules and capabilities. Amazon SageMaker Studio is the first integrated development environment (IDE) for machine learning. It gives users complete visibility and control over all the steps involved with building, training, and deploying machine learning models. All ML development activities including notebook creation, building models, training and experiment management, debugging, automatic model creation, and model deployment and monitoring can be performed within the unified SageMaker Studio visual interface, giving developers all the tools needed to build and manage machine learning models. Another key component to the service is Amazon SageMaker Ground Truth which helps users build, manage, and label training datasets for machine learning. The system is backed by machine learning so it learns from the labels created by humans to create automatic annotations which has the potential to substantially reduce data labeling costs by up to 70 percent. Customers Nucleus has spoken with reference the integration with Kubernetes as a source for potential time savings as users can train and deploy models in SageMaker using Kubernetes operators; the containerized machine learning models integrate across the Kubernetes ecosystem and often can be trained faster, allowing users to further reduce the costs and complexity of their machine learning programs.

• Cover